Logarithmic Sobolev inequalities in discrete product spaces: proof by a transportation cost distance

Size: px
Start display at page:

Download "Logarithmic Sobolev inequalities in discrete product spaces: proof by a transportation cost distance"

Transcription

1 Logarithmic Sobolev inequalities in discrete product spaces: proof by a transportation cost distance Katalin Marton Alfréd Rényi Institute of Mathematics of the Hungarian Academy of Sciences

2 Relative entropy Definition Z: measurable space, µ, ν: probability measures on Z. Z, U: random variables, L(Z) = µ, L(U) = ν. Relative entropy: D(µ ν) = D(Z U) = Z log dµ dµ (1) dν If Z is finite: D(µ ν) = D(Z U) = z Z µ(z) log µ(z) ν(z). (2)

3 Entropy contraction of Markov kernels Definition (Z, ν) : probability space, P(Z): measures on Z, Γ: Markov kernel on Z with invariant measure ν. Entropy contraction for (Z, ν, Γ) with rate 1 c, 0 < c 1: D(µΓ ν) (1 c) D(µ ν). (3) Equivalently: c D(µ ν) ( D(µ ν) D(µΓ ν) ), for all µ P(Z) (4)

4 Gibbs sampler governed by local specifications of q n Definition Γ i : Markov kernel X n X n Γ: Markov kernel X n X n : Definition Γ i (z n y n ) = δ(ȳ i, z i ) q i (z i ȳ i ), Γ = 1 n n Γ i. q n has the entropy contraction property if: its Gibbs sampler Γ has.

5 Entropy contraction in product space (X n, q n ): probability space Question Which measures q n have the entropy contraction property with a reasonable constant c? c cannot be larger than O(1/n). Changing notation: WHEN c n D(pn q n ) ( D(p n q n ) D(p n Γ q n ) ) for all p n P(X n )? (5) Equivalently: WHEN D(p n q n ) (1 c n ) D(pn Γ q n )? (6)

6 Conditional relative entropy Definition Z: measurable space, V: another measurable space, π: probability measure on V, V : random variable, L(V ) = π For v V probability measures on Z: µ( v) = L(Z V = v) ν( v) = L(U V = v) Conditional relative entropy: D ( µ( V ) ν( V ) ) = D ( Z V ) U V ) ) D ( µ( v) ν( v) ) dπ V (7)

7 Cont d If V is finite: D ( µ( V ) ν( V ) = D ( Z V ) U V ) ) = v V π(v)d ( µ( v) ν( v) ). (8)

8 Sufficient condition for entropy contraction in product space (X n, q n, Γ): probability space with Gibbs sampler, q n = L(X n ) p n = L(Y n ): another distribution on X n Recall: Γ = 1 n n Γ i Ȳ i = (Y 1,..., Y i 1, Y i+1,..., Y n ), Proposition c D(p n q n ) = n D ( ) p i ( Ȳi) q i ( Ȳi D(p n Γ q n ) ( 1 c n) D(p n q n ). If q n is a product measure then: c = 1. ( ) (9)

9 Notation For x n = (x 1, x 2,..., x n ) X n : x i (x 1,..., x i 1, x i+1,..., x n ) p i L(Ȳi), q i L( X i ) Expansion formula D(p n q n ) = D( p i q i ) + D ( p i ( Ȳi) q i ( Ȳi) ) = D(p n q n ) = 1 n n D( p i q i ) + 1 n n D ( p i ( Ȳi) q i ( Ȳi) )

10 Proof of Proposition D(p n q n ) D(p n Γ q n ) (by convexity of entropy) D(p n q n ) 1 n n D( p i q i ) = 1 n n D ( p i ( Ȳi q i ( Ȳi) ) (by assumption) c n D(pn q n )

11 Entropy contraction in DISCRETE product spaces X finite (X n, q n, Γ) Wanted: inequality n D(p n q n ) C D ( p i ( Ȳi q i ( Ȳi) ) for all p n P(X n ) ( ) To get (*): use a Wassersteine-like distance.

12 A reverse Pinsker s inequality µ, ν: probability measures on X X finite Notation: Variational distance µ ν = 1 2 Lemma Set Then x X µ(x) ν(x) X + {x X : ν(x) > 0}, α min { ν(x) : x X + } D(µ ν) 4 µ ν 2 α Follows from the inequality D(µ ν) x X µ(x) ν(x) 2 ν(x)

13 A distance between measures on product spaces X n : product space µ n = L(Z n ), ν n = L(U n ): probability measures on X n Definition (P. Massart) The square of the W 2 -distance of µ n and ν n : W2 2 (µ n, ν n ) min inf: on couplings of µ n and ν n. n P r 2{ } Z i U i

14 Notation For I [1, n], p n = L(Y n ) and y n X n : y I (y i : i I), ȳ I (y i : i / I) p I L(Y I ), p I ( ȳ I ) L(Y I ȲI = ȳ I ) For Theorem 1 we need the inequality W2 2 ( n p n q n ) C E p i ( Ȳi) q i ( Ȳi) 2. in a MORE GENERAL FORM: We require a bound for W 2 2 ( pi ( ȳ I ), q I ( ȳ I ) ) for ALL subsets I [1, n] (not just I = [1, n]), and all ȳ I.

15 Main Theorem (X n, q n ), X finite! Theorem 1 Set α = min { q i (x i x i ) : q n (x n ) > 0, 1 i n } (10) Fix a p n = L(Y n ) P(X n ); assume q n (x n ) = 0 = p n (x n ) = 0. (11) Main assumption: ( ) )) pi ( ȳi, qi ( ȳi W 2 2 { C E p i ( Ȳi) q i ( Ȳi) 2 Ȳ I = ȳ I }, i I (12) for all I [1, n], ȳ I X [1,n]\I.

16 Main Theorem continued Assume all the inequalities ( ) )) pi ( ȳi, qi ( ȳi W 2 2 { C E p i ( Ȳi) q i ( Ȳi) 2 Ȳ I = ȳ I }, i I (13) where I [1, n] and ȳ I X [1,n]\I is fixed. Then D(p n q n ) 4C n α E pi ( Ȳi) q i ( Ȳi) 2 2C α n D ( p i ( Ȳi) q i ( Ȳi) ). (14)

17 An analogous result for densities in R n Theorem f(x n ) = exp( V (x n ) : density on R n, q n : probability measure with density f. Assume: conditional densities a f(x i x i ) satisfy a logarithmic Sobolev inequality with constant ρ, for all i, x i. Under some conditions on 1 ρ Hess V (xn ) (expressing that V is not too far from being uniformly convex): D(p n q n ) C (C = C(q n )) n D ( p i ( Ȳi) q i ( Ȳi) ) for all p n

18 Proof of Theorem 1 By induction on n. Assume Theorem 1 for n 1 Notation p i ( y i ) L(Ȳi Y i = y i ) For every i [1, n] D(p n q n ) = D(Y n X n ) = D(Y i X i ) + D ( p i ( Y i ) q i ( Y i ) ) = D(p n q n ) = 1 n n D(Y i X i ) + 1 n n D ( p i ( Y i ) q i ( Ȳi) ) By the induction hypothesis the second term is (15) ( 1 1 n ) 4C α n E p i ( Ȳi) q i ( Ȳi) 2

19 Proof of Theorem 1 Cont d Second term: First term: 1 n ( 1 1 n n D(Y i X i ) ) 4C α n E p i ( Ȳi) q i ( Ȳi) 2 (by the Lemma) n P r 2{ } Y i X i in any coupling of p n, q n = 1 n 4 α W 2 2 (p n, q n ) for the best coupling (by the assumption of Theorem 1) 1 n 4C n α pi ( Ȳi) q i ( Ȳi) 2 1 n 4 α L(Y i ) L(X i ) 2 (16)

20 Entropy contraction (X n, q n, Γ), X finite Γ: Gibbs sampler Corollary 1 If q n satisfies the conditions of Theorem 1 then ( D(p n Γ q n ) 1 α ) D(p n q n ). (17) 2nC

21 Logarithmic Sobolev inequality Notation E: Dirichlet form associated with Γ is the quadratic form E(f, g) = (Id Γ)f, g q n Definition q n satisfies a logarithmic Sobolev inequality with constant c > 0 if: ( ) p c D(p n q n n p ) E q n, n q n for all p n P(X n )

22 Logarithmic Sobolev inequality for Gibbs sampler (X n, q n, Γ), X finite Corollary 2 Under conditions of Theorem 1 the logarithmic Sobolev inequality holds true: 1 n D(pn q n ) 4C ( ) p α E n p Γ q n, n q n = 4C n ( ( αn ( ( ) E 1 p i yi Ȳi) ) 2) qi yi Ȳi. y i X (18) = hypercontractivity

23 ***********************************?????????????????

24 Application: Gibbs measures with Dobrushin s uniqueness condition (X n, q n ), X finite Definition q n satisfies (an L 2 -version of) Dobrushin s uniqueness condition with coupling matrix if: A = ( a k,i ) n k,, a i,i = 0, (i) max q i ( z i ) q i ( s i ) a k,i, k i, max : for all z i, s i differing only in the k-th coordinate, (19) and (ii) A 2 < 1.

25 (X n, q n ), X finite Theorem 2 Assume Dobrushin s uniqueness condition with coupling matrix A, A 2 < 1. Then conditions of Theorem 1 hold with C = 1/ ( 1 A ) 2. Thus D(p n q n ) 4 α 2 α 1 ( 1 A ) 2 1 ( 1 A ) 2 n E p i ( Ȳi) q i ( Ȳi) 2 n D ( p i ( Ȳi) q i ( Ȳi) ), (20)

26 Cont d Proof Dobrushin s uniqueness condition implies that Γ contracts W 2 -distance with rate 1 1 ) (1 n A 2.

27 Application: Gibbs measures on Z d Notation Z d : d-dimensional lattice, i Z d : site ρ(k, i) = max 1 ν d k ν i ν : distance on Z d, Λ Z d : finite set of sites X finite: spin space x Zd = (x i : i Z d ) X Zd : spin configuration X Zd : configuration space, For x Zd and Λ Z d x Λ = (x i : i Λ), x Λ = (x i : i / Λ), x Λ is called an outside configuration for Λ.

28 Local specificatons Definition q Λ ( x Λ ), Λ Z d : conditional distributions on X Λ. Assume compatibility conditions. There exists at least one probability measure q on the space of configurations: q = L(X) P(X Zd ) satisfying L(X Λ X Λ = x Λ ) = q Λ ( x Λ ), all Λ Z d. q Λ ( x Λ ): local specifications of q.

29 Finite range interactions Definition The local specifications have finite range of interactions if: there is an R > 0: q Λ ( x Λ ) only depends on coordinates k / Λ with ρ(k, Λ) R. q may not be uniquely defined by the local specifications, even for finite range interactions.

30 Dobrushin-Shlosman s strong mixing condition Given local specifications q Λ ( x Λ ). Assumption There exists a function ϕ(ρ) of the distance such that: (i) ( ) ρ(k, i) <, and: (ii) for every i Z d ϕ Λ Z d, M Λ, k / Λ and every ȳ Λ, z Λ, differing only at k : qm ( ȳ Λ ) q M ( z Λ ) ( ) ϕ ρ(k, M).

31 In case of finite range interactions: If Dobrushin-Shlosman s strong mixing condition holds then ϕ(ρ) = C exp( γ ρ) can be taken

32 ρ(k, M) = max length of red segments q M ( y Λ ) q M ( z Λ ) ϕ(ρ(k, M)) y Λ, z Λ differ only at k Λ

33 Dobrushin-Shlosman s strong mixing condition Cont d Meaning: The influence of the spin at k / Λ on the spins in M Λ that are far away from k is small. Essential: The spins over Λ are fixed in two different ways. The spins over Λ \ M are not fixed.

34 Logarithmic Sobolev inequality for strongly mixing measures Earlier results for the case of finite range interactions: D. Stroock, B. Zegarlinski 1992, F. Cesi 2001 F. Martinelli, E. Olivieri Theorem 3 (X Λ, q Λ ( ȳ Λ )) for fixed Λ and ȳ Λ α = min { q i (x i x i ) : q i (x i x i ) > 0 }. (Finite range is not assumed.) { Dobrushin-Shlosman s strong mixing condition + {α > 0} } = conditions of Theorem 1 for q Λ ( ȳ Λ ), with uniform constant = logarithmic Sobolev inequality for q Λ ( ȳ Λ ) with uniform constant

35 Logarithmic Sobolev inequality for strongly mixing measures Cont d The proof uses a Gibbs sampler updating cubes of size m depending on the dimension and on the function ϕ(ρ). We get W 2 2 ( pλ, q Λ ( ȳ Λ ) ) C m I:m-sided cube E p I Λ ( ȲI Λ) q I Λ ( ȲI Λ) 2 C m,α E ) ) p i ( YΛ\i qi ( YΛ\i, ȳ Λ 2 i Λ for an appropriate m that is good enough for any Λ and ȳ Λ.

Distance-Divergence Inequalities

Distance-Divergence Inequalities Distance-Divergence Inequalities Katalin Marton Alfréd Rényi Institute of Mathematics of the Hungarian Academy of Sciences Motivation To find a simple proof of the Blowing-up Lemma, proved by Ahlswede,

More information

Stein s method, logarithmic Sobolev and transport inequalities

Stein s method, logarithmic Sobolev and transport inequalities Stein s method, logarithmic Sobolev and transport inequalities M. Ledoux University of Toulouse, France and Institut Universitaire de France Stein s method, logarithmic Sobolev and transport inequalities

More information

Displacement convexity of the relative entropy in the discrete h

Displacement convexity of the relative entropy in the discrete h Displacement convexity of the relative entropy in the discrete hypercube LAMA Université Paris Est Marne-la-Vallée Phenomena in high dimensions in geometric analysis, random matrices, and computational

More information

CUTOFF FOR THE ISING MODEL ON THE LATTICE

CUTOFF FOR THE ISING MODEL ON THE LATTICE CUTOFF FOR THE ISING MODEL ON THE LATTICE EYAL LUBETZKY AND ALLAN SLY Abstract. Introduced in 1963, Glauber dynamics is one of the most practiced and extensively studied methods for sampling the Ising

More information

Convex decay of Entropy for interacting systems

Convex decay of Entropy for interacting systems Convex decay of Entropy for interacting systems Paolo Dai Pra Università degli Studi di Padova Cambridge March 30, 2011 Paolo Dai Pra Convex decay of Entropy for interacting systems Cambridge, 03/11 1

More information

Uniqueness of Fokker-Planck equations for spin lattice systems (I): compact case

Uniqueness of Fokker-Planck equations for spin lattice systems (I): compact case Semigroup Forum (213) 86:583 591 DOI 1.17/s233-12-945-y RESEARCH ARTICLE Uniqueness of Fokker-Planck equations for spin lattice systems (I): compact case Ludovic Dan Lemle Ran Wang Liming Wu Received:

More information

Logarithmic Sobolev Inequalities

Logarithmic Sobolev Inequalities Logarithmic Sobolev Inequalities M. Ledoux Institut de Mathématiques de Toulouse, France logarithmic Sobolev inequalities what they are, some history analytic, geometric, optimal transportation proofs

More information

Refined Bounds on the Empirical Distribution of Good Channel Codes via Concentration Inequalities

Refined Bounds on the Empirical Distribution of Good Channel Codes via Concentration Inequalities Refined Bounds on the Empirical Distribution of Good Channel Codes via Concentration Inequalities Maxim Raginsky and Igal Sason ISIT 2013, Istanbul, Turkey Capacity-Achieving Channel Codes The set-up DMC

More information

Combinatorial Criteria for Uniqueness of Gibbs Measures

Combinatorial Criteria for Uniqueness of Gibbs Measures Combinatorial Criteria for Uniqueness of Gibbs Measures Dror Weitz School of Mathematics, Institute for Advanced Study, Princeton, NJ 08540, U.S.A. dror@ias.edu April 12, 2005 Abstract We generalize previously

More information

Combinatorial Criteria for Uniqueness of Gibbs Measures

Combinatorial Criteria for Uniqueness of Gibbs Measures Combinatorial Criteria for Uniqueness of Gibbs Measures Dror Weitz* School of Mathematics, Institute for Advanced Study, Princeton, New Jersey 08540; e-mail: dror@ias.edu Received 17 November 2003; accepted

More information

Heat Flows, Geometric and Functional Inequalities

Heat Flows, Geometric and Functional Inequalities Heat Flows, Geometric and Functional Inequalities M. Ledoux Institut de Mathématiques de Toulouse, France heat flow and semigroup interpolations Duhamel formula (19th century) pde, probability, dynamics

More information

Solution Sheet 3. Solution Consider. with the metric. We also define a subset. and thus for any x, y X 0

Solution Sheet 3. Solution Consider. with the metric. We also define a subset. and thus for any x, y X 0 Solution Sheet Throughout this sheet denotes a domain of R n with sufficiently smooth boundary. 1. Let 1 p

More information

Concentration inequalities: basics and some new challenges

Concentration inequalities: basics and some new challenges Concentration inequalities: basics and some new challenges M. Ledoux University of Toulouse, France & Institut Universitaire de France Measure concentration geometric functional analysis, probability theory,

More information

Exercises Measure Theoretic Probability

Exercises Measure Theoretic Probability Exercises Measure Theoretic Probability 2002-2003 Week 1 1. Prove the folloing statements. (a) The intersection of an arbitrary family of d-systems is again a d- system. (b) The intersection of an arbitrary

More information

Concentration Inequalities for Dependent Random Variables via the Martingale Method

Concentration Inequalities for Dependent Random Variables via the Martingale Method Concentration Inequalities for Dependent Random Variables via the Martingale Method Leonid Kontorovich, and Kavita Ramanan Carnegie Mellon University L. Kontorovich School of Computer Science Carnegie

More information

Contents 1. Introduction 1 2. Main results 3 3. Proof of the main inequalities 7 4. Application to random dynamical systems 11 References 16

Contents 1. Introduction 1 2. Main results 3 3. Proof of the main inequalities 7 4. Application to random dynamical systems 11 References 16 WEIGHTED CSISZÁR-KULLBACK-PINSKER INEQUALITIES AND APPLICATIONS TO TRANSPORTATION INEQUALITIES FRANÇOIS BOLLEY AND CÉDRIC VILLANI Abstract. We strengthen the usual Csiszár-Kullback-Pinsker inequality by

More information

Stability results for Logarithmic Sobolev inequality

Stability results for Logarithmic Sobolev inequality Stability results for Logarithmic Sobolev inequality Daesung Kim (joint work with Emanuel Indrei) Department of Mathematics Purdue University September 20, 2017 Daesung Kim (Purdue) Stability for LSI Probability

More information

Recall that if X is a compact metric space, C(X), the space of continuous (real-valued) functions on X, is a Banach space with the norm

Recall that if X is a compact metric space, C(X), the space of continuous (real-valued) functions on X, is a Banach space with the norm Chapter 13 Radon Measures Recall that if X is a compact metric space, C(X), the space of continuous (real-valued) functions on X, is a Banach space with the norm (13.1) f = sup x X f(x). We want to identify

More information

curvature, mixing, and entropic interpolation Simons Feb-2016 and CSE 599s Lecture 13

curvature, mixing, and entropic interpolation Simons Feb-2016 and CSE 599s Lecture 13 curvature, mixing, and entropic interpolation Simons Feb-2016 and CSE 599s Lecture 13 James R. Lee University of Washington Joint with Ronen Eldan (Weizmann) and Joseph Lehec (Paris-Dauphine) Markov chain

More information

Ergodic Theorems. Samy Tindel. Purdue University. Probability Theory 2 - MA 539. Taken from Probability: Theory and examples by R.

Ergodic Theorems. Samy Tindel. Purdue University. Probability Theory 2 - MA 539. Taken from Probability: Theory and examples by R. Ergodic Theorems Samy Tindel Purdue University Probability Theory 2 - MA 539 Taken from Probability: Theory and examples by R. Durrett Samy T. Ergodic theorems Probability Theory 1 / 92 Outline 1 Definitions

More information

. Then V l on K l, and so. e e 1.

. Then V l on K l, and so. e e 1. Sanov s Theorem Let E be a Polish space, and define L n : E n M E to be the empirical measure given by L n x = n n m= δ x m for x = x,..., x n E n. Given a µ M E, denote by µ n the distribution of L n

More information

CUTOFF FOR GENERAL SPIN SYSTEMS WITH ARBITRARY BOUNDARY CONDITIONS

CUTOFF FOR GENERAL SPIN SYSTEMS WITH ARBITRARY BOUNDARY CONDITIONS CUTOFF FOR GENERAL SPIN SYSTEMS WITH ARBITRARY BOUNDARY CONDITIONS EYAL LUBETZKY AND ALLAN SLY Abstract. The cutoff phenomenon describes a sharp transition in the convergence of a Markov chain to equilibrium.

More information

Chapter 9: Basic of Hypercontractivity

Chapter 9: Basic of Hypercontractivity Analysis of Boolean Functions Prof. Ryan O Donnell Chapter 9: Basic of Hypercontractivity Date: May 6, 2017 Student: Chi-Ning Chou Index Problem Progress 1 Exercise 9.3 (Tightness of Bonami Lemma) 2/2

More information

Metric Spaces and Topology

Metric Spaces and Topology Chapter 2 Metric Spaces and Topology From an engineering perspective, the most important way to construct a topology on a set is to define the topology in terms of a metric on the set. This approach underlies

More information

Mathematics for Economists

Mathematics for Economists Mathematics for Economists Victor Filipe Sao Paulo School of Economics FGV Metric Spaces: Basic Definitions Victor Filipe (EESP/FGV) Mathematics for Economists Jan.-Feb. 2017 1 / 34 Definitions and Examples

More information

A simple condition implying rapid mixing of single-site dynamics on spin systems

A simple condition implying rapid mixing of single-site dynamics on spin systems A simple condition implying rapid mixing of single-site dynamics on spin systems Thomas P. Hayes University of California Berkeley Computer Science Division hayest@cs.berkeley.edu Abstract Spin systems

More information

Hypercontractivity of spherical averages in Hamming space

Hypercontractivity of spherical averages in Hamming space Hypercontractivity of spherical averages in Hamming space Yury Polyanskiy Department of EECS MIT yp@mit.edu Allerton Conference, Oct 2, 2014 Yury Polyanskiy Hypercontractivity in Hamming space 1 Hypercontractivity:

More information

Convergence to equilibrium of Markov processes (eventually piecewise deterministic)

Convergence to equilibrium of Markov processes (eventually piecewise deterministic) Convergence to equilibrium of Markov processes (eventually piecewise deterministic) A. Guillin Université Blaise Pascal and IUF Rennes joint works with D. Bakry, F. Barthe, F. Bolley, P. Cattiaux, R. Douc,

More information

MATH41011/MATH61011: FOURIER SERIES AND LEBESGUE INTEGRATION. Extra Reading Material for Level 4 and Level 6

MATH41011/MATH61011: FOURIER SERIES AND LEBESGUE INTEGRATION. Extra Reading Material for Level 4 and Level 6 MATH41011/MATH61011: FOURIER SERIES AND LEBESGUE INTEGRATION Extra Reading Material for Level 4 and Level 6 Part A: Construction of Lebesgue Measure The first part the extra material consists of the construction

More information

MONOTONE COUPLING AND THE ISING MODEL

MONOTONE COUPLING AND THE ISING MODEL MONOTONE COUPLING AND THE ISING MODEL 1. PERFECT MATCHING IN BIPARTITE GRAPHS Definition 1. A bipartite graph is a graph G = (V, E) whose vertex set V can be partitioned into two disjoint set V I, V O

More information

NEW FUNCTIONAL INEQUALITIES

NEW FUNCTIONAL INEQUALITIES 1 / 29 NEW FUNCTIONAL INEQUALITIES VIA STEIN S METHOD Giovanni Peccati (Luxembourg University) IMA, Minneapolis: April 28, 2015 2 / 29 INTRODUCTION Based on two joint works: (1) Nourdin, Peccati and Swan

More information

Generalized Orlicz spaces and Wasserstein distances for convex concave scale functions

Generalized Orlicz spaces and Wasserstein distances for convex concave scale functions Bull. Sci. math. 135 (2011 795 802 www.elsevier.com/locate/bulsci Generalized Orlicz spaces and Wasserstein distances for convex concave scale functions Karl-Theodor Sturm Institut für Angewandte Mathematik,

More information

ON THE ZERO-ONE LAW AND THE LAW OF LARGE NUMBERS FOR RANDOM WALK IN MIXING RAN- DOM ENVIRONMENT

ON THE ZERO-ONE LAW AND THE LAW OF LARGE NUMBERS FOR RANDOM WALK IN MIXING RAN- DOM ENVIRONMENT Elect. Comm. in Probab. 10 (2005), 36 44 ELECTRONIC COMMUNICATIONS in PROBABILITY ON THE ZERO-ONE LAW AND THE LAW OF LARGE NUMBERS FOR RANDOM WALK IN MIXING RAN- DOM ENVIRONMENT FIRAS RASSOUL AGHA Department

More information

1 Math 241A-B Homework Problem List for F2015 and W2016

1 Math 241A-B Homework Problem List for F2015 and W2016 1 Math 241A-B Homework Problem List for F2015 W2016 1.1 Homework 1. Due Wednesday, October 7, 2015 Notation 1.1 Let U be any set, g be a positive function on U, Y be a normed space. For any f : U Y let

More information

Lecture 7. µ(x)f(x). When µ is a probability measure, we say µ is a stationary distribution.

Lecture 7. µ(x)f(x). When µ is a probability measure, we say µ is a stationary distribution. Lecture 7 1 Stationary measures of a Markov chain We now study the long time behavior of a Markov Chain: in particular, the existence and uniqueness of stationary measures, and the convergence of the distribution

More information

ESTIMATES FOR THE MONGE-AMPERE EQUATION

ESTIMATES FOR THE MONGE-AMPERE EQUATION GLOBAL W 2,p ESTIMATES FOR THE MONGE-AMPERE EQUATION O. SAVIN Abstract. We use a localization property of boundary sections for solutions to the Monge-Ampere equation obtain global W 2,p estimates under

More information

Capacitary inequalities in discrete setting and application to metastable Markov chains. André Schlichting

Capacitary inequalities in discrete setting and application to metastable Markov chains. André Schlichting Capacitary inequalities in discrete setting and application to metastable Markov chains André Schlichting Institute for Applied Mathematics, University of Bonn joint work with M. Slowik (TU Berlin) 12

More information

An example of application of discrete Hardy s inequalities

An example of application of discrete Hardy s inequalities An example of application of discrete Hardy s inequalities Laurent Miclo Laboratoire de Statistique et Probabilités, UMR C5583 Université PaulSabatieretCNRS 8, route de Narbonne 3062 Toulouse cedex, France

More information

THEOREMS, ETC., FOR MATH 516

THEOREMS, ETC., FOR MATH 516 THEOREMS, ETC., FOR MATH 516 Results labeled Theorem Ea.b.c (or Proposition Ea.b.c, etc.) refer to Theorem c from section a.b of Evans book (Partial Differential Equations). Proposition 1 (=Proposition

More information

MATH642. COMPLEMENTS TO INTRODUCTION TO DYNAMICAL SYSTEMS BY M. BRIN AND G. STUCK

MATH642. COMPLEMENTS TO INTRODUCTION TO DYNAMICAL SYSTEMS BY M. BRIN AND G. STUCK MATH642 COMPLEMENTS TO INTRODUCTION TO DYNAMICAL SYSTEMS BY M BRIN AND G STUCK DMITRY DOLGOPYAT 13 Expanding Endomorphisms of the circle Let E 10 : S 1 S 1 be given by E 10 (x) = 10x mod 1 Exercise 1 Show

More information

Example (cont d): Function fields in one variable...

Example (cont d): Function fields in one variable... Example (cont d): Function fields in one variable... Garrett 10-17-2011 1 Practice: consider K a finite extension of k = C(X), and O the integral closure in K of o = C[X]. K = C(X, Y ) for some Y, and

More information

Discrete Ricci curvature via convexity of the entropy

Discrete Ricci curvature via convexity of the entropy Discrete Ricci curvature via convexity of the entropy Jan Maas University of Bonn Joint work with Matthias Erbar Simons Institute for the Theory of Computing UC Berkeley 2 October 2013 Starting point McCann

More information

Rigorous approximation of invariant measures for IFS Joint work April 8, with 2016 S. 1 Galat / 21

Rigorous approximation of invariant measures for IFS Joint work April 8, with 2016 S. 1 Galat / 21 Rigorous approximation of invariant measures for IFS Joint work with S. Galatolo e I. Nisoli Maurizio Monge maurizio.monge@im.ufrj.br Universidade Federal do Rio de Janeiro April 8, 2016 Rigorous approximation

More information

Exercises Measure Theoretic Probability

Exercises Measure Theoretic Probability Exercises Measure Theoretic Probability Chapter 1 1. Prove the folloing statements. (a) The intersection of an arbitrary family of d-systems is again a d- system. (b) The intersection of an arbitrary family

More information

Course 212: Academic Year Section 1: Metric Spaces

Course 212: Academic Year Section 1: Metric Spaces Course 212: Academic Year 1991-2 Section 1: Metric Spaces D. R. Wilkins Contents 1 Metric Spaces 3 1.1 Distance Functions and Metric Spaces............. 3 1.2 Convergence and Continuity in Metric Spaces.........

More information

Invariances in spectral estimates. Paris-Est Marne-la-Vallée, January 2011

Invariances in spectral estimates. Paris-Est Marne-la-Vallée, January 2011 Invariances in spectral estimates Franck Barthe Dario Cordero-Erausquin Paris-Est Marne-la-Vallée, January 2011 Notation Notation Given a probability measure ν on some Euclidean space, the Poincaré constant

More information

An Inverse Problem for Gibbs Fields with Hard Core Potential

An Inverse Problem for Gibbs Fields with Hard Core Potential An Inverse Problem for Gibbs Fields with Hard Core Potential Leonid Koralov Department of Mathematics University of Maryland College Park, MD 20742-4015 koralov@math.umd.edu Abstract It is well known that

More information

An upper bound on l q norms of noisy functions

An upper bound on l q norms of noisy functions Electronic Collouium on Computational Complexity, Report No. 68 08 An upper bound on l norms of noisy functions Alex Samorodnitsky Abstract Let T ɛ, 0 ɛ /, be the noise operator acting on functions on

More information

DETECTING PHASE TRANSITION FOR GIBBS MEASURES. By Francis Comets 1 University of California, Irvine

DETECTING PHASE TRANSITION FOR GIBBS MEASURES. By Francis Comets 1 University of California, Irvine The Annals of Applied Probability 1997, Vol. 7, No. 2, 545 563 DETECTING PHASE TRANSITION FOR GIBBS MEASURES By Francis Comets 1 University of California, Irvine We propose a new empirical procedure for

More information

arxiv: v1 [math.fa] 1 Oct 2018

arxiv: v1 [math.fa] 1 Oct 2018 GRADIENT FLOW STRUCTURE AND EXPONENTIAL DECAY OF THE SANDWICHED RÉNYI DIVERGENCE FOR PRIMITIVE LINDBLAD EQUATIONS WITH GNS-DETAILED BALANCE arxiv:18196v1 [mathfa] 1 Oct 18 YU CAO, JIANFENG LU, AND YULONG

More information

Fokker-Planck Equation on Graph with Finite Vertices

Fokker-Planck Equation on Graph with Finite Vertices Fokker-Planck Equation on Graph with Finite Vertices January 13, 2011 Jointly with S-N Chow (Georgia Tech) Wen Huang (USTC) Hao-min Zhou(Georgia Tech) Functional Inequalities and Discrete Spaces Outline

More information

Walsh Diffusions. Andrey Sarantsev. March 27, University of California, Santa Barbara. Andrey Sarantsev University of Washington, Seattle 1 / 1

Walsh Diffusions. Andrey Sarantsev. March 27, University of California, Santa Barbara. Andrey Sarantsev University of Washington, Seattle 1 / 1 Walsh Diffusions Andrey Sarantsev University of California, Santa Barbara March 27, 2017 Andrey Sarantsev University of Washington, Seattle 1 / 1 Walsh Brownian Motion on R d Spinning measure µ: probability

More information

LECTURE 10: THE ATIYAH-GUILLEMIN-STERNBERG CONVEXITY THEOREM

LECTURE 10: THE ATIYAH-GUILLEMIN-STERNBERG CONVEXITY THEOREM LECTURE 10: THE ATIYAH-GUILLEMIN-STERNBERG CONVEXITY THEOREM Contents 1. The Atiyah-Guillemin-Sternberg Convexity Theorem 1 2. Proof of the Atiyah-Guillemin-Sternberg Convexity theorem 3 3. Morse theory

More information

Concentration inequalities and the entropy method

Concentration inequalities and the entropy method Concentration inequalities and the entropy method Gábor Lugosi ICREA and Pompeu Fabra University Barcelona what is concentration? We are interested in bounding random fluctuations of functions of many

More information

Stochastic relations of random variables and processes

Stochastic relations of random variables and processes Stochastic relations of random variables and processes Lasse Leskelä Helsinki University of Technology 7th World Congress in Probability and Statistics Singapore, 18 July 2008 Fundamental problem of applied

More information

Convergence to equilibrium for rough differential equations

Convergence to equilibrium for rough differential equations Convergence to equilibrium for rough differential equations Samy Tindel Purdue University Barcelona GSE Summer Forum 2017 Joint work with Aurélien Deya (Nancy) and Fabien Panloup (Angers) Samy T. (Purdue)

More information

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation POINTWISE C 2,α ESTIMATES AT THE BOUNDARY FOR THE MONGE-AMPERE EQUATION O. SAVIN Abstract. We prove a localization property of boundary sections for solutions to the Monge-Ampere equation. As a consequence

More information

Defining the Integral

Defining the Integral Defining the Integral In these notes we provide a careful definition of the Lebesgue integral and we prove each of the three main convergence theorems. For the duration of these notes, let (, M, µ) be

More information

Path Coupling and Aggregate Path Coupling

Path Coupling and Aggregate Path Coupling Path Coupling and Aggregate Path Coupling 0 Path Coupling and Aggregate Path Coupling Yevgeniy Kovchegov Oregon State University (joint work with Peter T. Otto from Willamette University) Path Coupling

More information

Can extra updates delay mixing?

Can extra updates delay mixing? Can extra updates delay mixing? Yuval Peres Peter Winkler July 7, 2010 Abstract We consider Glauber dynamics (starting from an extremal configuration) in a monotone spin system, and show that interjecting

More information

Sobolev Spaces. Chapter 10

Sobolev Spaces. Chapter 10 Chapter 1 Sobolev Spaces We now define spaces H 1,p (R n ), known as Sobolev spaces. For u to belong to H 1,p (R n ), we require that u L p (R n ) and that u have weak derivatives of first order in L p

More information

OXPORD UNIVERSITY PRESS

OXPORD UNIVERSITY PRESS Concentration Inequalities A Nonasymptotic Theory of Independence STEPHANE BOUCHERON GABOR LUGOSI PASCAL MASS ART OXPORD UNIVERSITY PRESS CONTENTS 1 Introduction 1 1.1 Sums of Independent Random Variables

More information

VARIATIONAL PRINCIPLE FOR THE ENTROPY

VARIATIONAL PRINCIPLE FOR THE ENTROPY VARIATIONAL PRINCIPLE FOR THE ENTROPY LUCIAN RADU. Metric entropy Let (X, B, µ a measure space and I a countable family of indices. Definition. We say that ξ = {C i : i I} B is a measurable partition if:

More information

Math 5051 Measure Theory and Functional Analysis I Homework Assignment 2

Math 5051 Measure Theory and Functional Analysis I Homework Assignment 2 Math 551 Measure Theory and Functional nalysis I Homework ssignment 2 Prof. Wickerhauser Due Friday, September 25th, 215 Please do Exercises 1, 4*, 7, 9*, 11, 12, 13, 16, 21*, 26, 28, 31, 32, 33, 36, 37.

More information

Concentration Properties of Restricted Measures with Applications to Non-Lipschitz Functions

Concentration Properties of Restricted Measures with Applications to Non-Lipschitz Functions Concentration Properties of Restricted Measures with Applications to Non-Lipschitz Functions S G Bobkov, P Nayar, and P Tetali April 4, 6 Mathematics Subject Classification Primary 6Gxx Keywords and phrases

More information

Myopic Models of Population Dynamics on Infinite Networks

Myopic Models of Population Dynamics on Infinite Networks Myopic Models of Population Dynamics on Infinite Networks Robert Carlson Department of Mathematics University of Colorado at Colorado Springs rcarlson@uccs.edu June 30, 2014 Outline Reaction-diffusion

More information

Peacocks nearby: approximating sequences of measures

Peacocks nearby: approximating sequences of measures Peacocks nearby: approximating sequences of measures Stefan Gerhold TU Wien sgerhold@fam.tuwien.ac.at I. Cetin Gülüm TU Wien ismail.cetin.gueluem@gmx.at October 2, 2017 Abstract A peacock is a family of

More information

KLS-TYPE ISOPERIMETRIC BOUNDS FOR LOG-CONCAVE PROBABILITY MEASURES. December, 2014

KLS-TYPE ISOPERIMETRIC BOUNDS FOR LOG-CONCAVE PROBABILITY MEASURES. December, 2014 KLS-TYPE ISOPERIMETRIC BOUNDS FOR LOG-CONCAVE PROBABILITY MEASURES Sergey G. Bobkov and Dario Cordero-Erausquin December, 04 Abstract The paper considers geometric lower bounds on the isoperimetric constant

More information

ON THE ISING MODEL WITH RANDOM BOUNDARY CONDITION

ON THE ISING MODEL WITH RANDOM BOUNDARY CONDITION ON THE ISING MODEL WITH RANDOM BOUNDARY CONDITION A. C. D. VAN ENTER, K. NETOČNÝ, AND H. G. SCHAAP ABSTRACT. The infinite-volume limit behavior of the 2d Ising model under possibly strong random boundary

More information

University of Toronto Department of Statistics

University of Toronto Department of Statistics Norm Comparisons for Data Augmentation by James P. Hobert Department of Statistics University of Florida and Jeffrey S. Rosenthal Department of Statistics University of Toronto Technical Report No. 0704

More information

Lecture 3: Expected Value. These integrals are taken over all of Ω. If we wish to integrate over a measurable subset A Ω, we will write

Lecture 3: Expected Value. These integrals are taken over all of Ω. If we wish to integrate over a measurable subset A Ω, we will write Lecture 3: Expected Value 1.) Definitions. If X 0 is a random variable on (Ω, F, P), then we define its expected value to be EX = XdP. Notice that this quantity may be. For general X, we say that EX exists

More information

Cauchy Duals of 2-hyperexpansive Operators. The Multivariable Case

Cauchy Duals of 2-hyperexpansive Operators. The Multivariable Case Operators Cauchy Dual to 2-hyperexpansive Operators: The Multivariable Case Raúl E. Curto Southeastern Analysis Meeting XXVII, Gainesville (joint work with Sameer Chavan) raul-curto@uiowa.edu http://www.math.uiowa.edu/

More information

Measurable functions are approximately nice, even if look terrible.

Measurable functions are approximately nice, even if look terrible. Tel Aviv University, 2015 Functions of real variables 74 7 Approximation 7a A terrible integrable function........... 74 7b Approximation of sets................ 76 7c Approximation of functions............

More information

arxiv: v2 [math.ds] 24 Apr 2018

arxiv: v2 [math.ds] 24 Apr 2018 CONSTRUCTION OF SOME CHOWLA SEQUENCES RUXI SHI arxiv:1804.03851v2 [math.ds] 24 Apr 2018 Abstract. For numerical sequences taking values 0 or complex numbers of modulus 1, we define Chowla property and

More information

Part II Probability and Measure

Part II Probability and Measure Part II Probability and Measure Theorems Based on lectures by J. Miller Notes taken by Dexter Chua Michaelmas 2016 These notes are not endorsed by the lecturers, and I have modified them (often significantly)

More information

Convexity, Detection, and Generalized f-divergences

Convexity, Detection, and Generalized f-divergences Convexity, Detection, and Generalized f-divergences Khashayar Khosravi Feng Ruan John Duchi June 5, 015 1 Introduction The goal of classification problem is to learn a discriminant function for classification

More information

arxiv:math/ v1 [math.pr] 21 Aug 2006

arxiv:math/ v1 [math.pr] 21 Aug 2006 Measure Concentration of Markov Tree Processes arxiv:math/0608511v1 [math.pr] 21 Aug 2006 Leonid Kontorovich School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 USA lkontor@cs.cmu.edu

More information

ON LARGE DEVIATIONS OF MARKOV PROCESSES WITH DISCONTINUOUS STATISTICS 1

ON LARGE DEVIATIONS OF MARKOV PROCESSES WITH DISCONTINUOUS STATISTICS 1 The Annals of Applied Probability 1998, Vol. 8, No. 1, 45 66 ON LARGE DEVIATIONS OF MARKOV PROCESSES WITH DISCONTINUOUS STATISTICS 1 By Murat Alanyali 2 and Bruce Hajek Bell Laboratories and University

More information

Weak convergence in Probability Theory A summer excursion! Day 3

Weak convergence in Probability Theory A summer excursion! Day 3 BCAM June 2013 1 Weak convergence in Probability Theory A summer excursion! Day 3 Armand M. Makowski ECE & ISR/HyNet University of Maryland at College Park armand@isr.umd.edu BCAM June 2013 2 Day 1: Basic

More information

Continuity. Matt Rosenzweig

Continuity. Matt Rosenzweig Continuity Matt Rosenzweig Contents 1 Continuity 1 1.1 Rudin Chapter 4 Exercises........................................ 1 1.1.1 Exercise 1............................................. 1 1.1.2 Exercise

More information

A note on the convex infimum convolution inequality

A note on the convex infimum convolution inequality A note on the convex infimum convolution inequality Naomi Feldheim, Arnaud Marsiglietti, Piotr Nayar, Jing Wang Abstract We characterize the symmetric measures which satisfy the one dimensional convex

More information

Real Analysis Math 131AH Rudin, Chapter #1. Dominique Abdi

Real Analysis Math 131AH Rudin, Chapter #1. Dominique Abdi Real Analysis Math 3AH Rudin, Chapter # Dominique Abdi.. If r is rational (r 0) and x is irrational, prove that r + x and rx are irrational. Solution. Assume the contrary, that r+x and rx are rational.

More information

Mathematical Institute, University of Utrecht. The problem of estimating the mean of an observed Gaussian innite-dimensional vector

Mathematical Institute, University of Utrecht. The problem of estimating the mean of an observed Gaussian innite-dimensional vector On Minimax Filtering over Ellipsoids Eduard N. Belitser and Boris Y. Levit Mathematical Institute, University of Utrecht Budapestlaan 6, 3584 CD Utrecht, The Netherlands The problem of estimating the mean

More information

THEOREMS, ETC., FOR MATH 515

THEOREMS, ETC., FOR MATH 515 THEOREMS, ETC., FOR MATH 515 Proposition 1 (=comment on page 17). If A is an algebra, then any finite union or finite intersection of sets in A is also in A. Proposition 2 (=Proposition 1.1). For every

More information

p 1 ( Y p dp) 1/p ( X p dp) 1 1 p

p 1 ( Y p dp) 1/p ( X p dp) 1 1 p Doob s inequality Let X(t) be a right continuous submartingale with respect to F(t), t 1 P(sup s t X(s) λ) 1 λ {sup s t X(s) λ} X + (t)dp 2 For 1 < p

More information

215 Problem 1. (a) Define the total variation distance µ ν tv for probability distributions µ, ν on a finite set S. Show that

215 Problem 1. (a) Define the total variation distance µ ν tv for probability distributions µ, ν on a finite set S. Show that 15 Problem 1. (a) Define the total variation distance µ ν tv for probability distributions µ, ν on a finite set S. Show that µ ν tv = (1/) x S µ(x) ν(x) = x S(µ(x) ν(x)) + where a + = max(a, 0). Show that

More information

ON THE DEFINITION OF RELATIVE PRESSURE FOR FACTOR MAPS ON SHIFTS OF FINITE TYPE. 1. Introduction

ON THE DEFINITION OF RELATIVE PRESSURE FOR FACTOR MAPS ON SHIFTS OF FINITE TYPE. 1. Introduction ON THE DEFINITION OF RELATIVE PRESSURE FOR FACTOR MAPS ON SHIFTS OF FINITE TYPE KARL PETERSEN AND SUJIN SHIN Abstract. We show that two natural definitions of the relative pressure function for a locally

More information

INTRODUCTION TO STATISTICAL MECHANICS Exercises

INTRODUCTION TO STATISTICAL MECHANICS Exercises École d hiver de Probabilités Semaine 1 : Mécanique statistique de l équilibre CIRM - Marseille 4-8 Février 2012 INTRODUCTION TO STATISTICAL MECHANICS Exercises Course : Yvan Velenik, Exercises : Loren

More information

Lorenz like flows. Maria José Pacifico. IM-UFRJ Rio de Janeiro - Brasil. Lorenz like flows p. 1

Lorenz like flows. Maria José Pacifico. IM-UFRJ Rio de Janeiro - Brasil. Lorenz like flows p. 1 Lorenz like flows Maria José Pacifico pacifico@im.ufrj.br IM-UFRJ Rio de Janeiro - Brasil Lorenz like flows p. 1 Main goals The main goal is to explain the results (Galatolo-P) Theorem A. (decay of correlation

More information

On Approximate Pattern Matching for a Class of Gibbs Random Fields

On Approximate Pattern Matching for a Class of Gibbs Random Fields On Approximate Pattern Matching for a Class of Gibbs Random Fields J.-R. Chazottes F. Redig E. Verbitskiy March 3, 2005 Abstract: We prove an exponential approximation for the law of approximate occurrence

More information

ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS

ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS Bendikov, A. and Saloff-Coste, L. Osaka J. Math. 4 (5), 677 7 ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS ALEXANDER BENDIKOV and LAURENT SALOFF-COSTE (Received March 4, 4)

More information

Dynamics for the critical 2D Potts/FK model: many questions and a few answers

Dynamics for the critical 2D Potts/FK model: many questions and a few answers Dynamics for the critical 2D Potts/FK model: many questions and a few answers Eyal Lubetzky May 2018 Courant Institute, New York University Outline The models: static and dynamical Dynamical phase transitions

More information

Random sampling for the monomer-dimer model on a lattice

Random sampling for the monomer-dimer model on a lattice Centrum voor Wiskunde en Informatica Random sampling for the monomer-dimer model on a lattice J. van den Berg, R. Brouwer Probability, Networks and Algorithms (PNA) PNA-R9917 December 31, 1999 Report PNA-R9917

More information

The Gaussian free field, Gibbs measures and NLS on planar domains

The Gaussian free field, Gibbs measures and NLS on planar domains The Gaussian free field, Gibbs measures and on planar domains N. Burq, joint with L. Thomann (Nantes) and N. Tzvetkov (Cergy) Université Paris Sud, Laboratoire de Mathématiques d Orsay, CNRS UMR 8628 LAGA,

More information

Information theoretic perspectives on learning algorithms

Information theoretic perspectives on learning algorithms Information theoretic perspectives on learning algorithms Varun Jog University of Wisconsin - Madison Departments of ECE and Mathematics Shannon Channel Hangout! May 8, 2018 Jointly with Adrian Tovar-Lopez

More information

Value and Policy Iteration

Value and Policy Iteration Chapter 7 Value and Policy Iteration 1 For infinite horizon problems, we need to replace our basic computational tool, the DP algorithm, which we used to compute the optimal cost and policy for finite

More information

Lattice spin models: Crash course

Lattice spin models: Crash course Chapter 1 Lattice spin models: Crash course 1.1 Basic setup Here we will discuss the basic setup of the models to which we will direct our attention throughout this course. The basic ingredients are as

More information

The Moment Method; Convex Duality; and Large/Medium/Small Deviations

The Moment Method; Convex Duality; and Large/Medium/Small Deviations Stat 928: Statistical Learning Theory Lecture: 5 The Moment Method; Convex Duality; and Large/Medium/Small Deviations Instructor: Sham Kakade The Exponential Inequality and Convex Duality The exponential

More information

A Common Fixed Points in Cone and rectangular cone Metric Spaces

A Common Fixed Points in Cone and rectangular cone Metric Spaces Chapter 5 A Common Fixed Points in Cone and rectangular cone Metric Spaces In this chapter we have establish fixed point theorems in cone metric spaces and rectangular cone metric space. In the first part

More information